[Interest] USB support

Konrad Rosenbaum konrad at silmor.de
Wed Apr 18 11:32:56 CEST 2018


On Tue, April 17, 2018 16:39, Roland Hughes wrote:
> On 04/17/2018 09:15 AM, Thiago Macieira wrote:
>> On Monday, 16 April 2018 17:16:43 PDT Roland Hughes wrote:
>>> I know the answer may well be RTFD, but, has raw USB communications
>>> been
>>> more integrated or are the libusb examples from 2008 "current"?
>> Nothing changed since 2008. You need to run as root anyway, so it's not
>> a
>> functionality you should expect to see in Qt any time soon.
> Interesting.
> I don't see any mention of needing to be root here:

> And the current "official" site http://libusb.info
> It is user-mode: No special privilege or elevation is required for the
> application to communicate with a device.

You have 2 options: a) be root, b) setup a udev rule to make the USB
device accessible to other users.

I leave the task of guessing which one is recommended and which one
happens most often in practice up to you as a practice. ;-)

> Of course, root, for an embedded system, isn't much of an issue.

Yes, it is. You can destroy the device as root. UIs or complicated
calculations should never run as root.

> One
> would think that being a member of plugdev would give you USB access
> much like being a member of dialout gives you access to serial port.

One would think. But one would often be wrong.

> It does appear there should be some kind of support already there if one
> built the webengine

Just because WebEngine uses it does not mean you get a public API from
there. It actually means there is potential trouble since you may be
running with 2 instances of libusb.

>       Can I run libusb applications on Linux without root privilege?
> Yes.
> The standard solution is to use udev rules. Here are some links to udev
> related websites.

Correct. It is pretty easy, once you've read a few of those files.

> Should prove to be an interesting experiment. I imagine a few dozen
> people have written their own QObject based wrapper class for that
> library but getting code contributed is monumentally impossible so they
> are all one-offs.

...and they are likely very project dependent.


More information about the Interest mailing list